1_mw_sleep
Jackie
05/31/21; 6/11/21; 7/4/21; 7/26/21
This script wrangles MW subjctive sleep data and computes metrics (e.g., RMSSD) for further analysis
1 loading libraries
library(tidyverse)
library(foreign)
library(haven)
library(readxl)
library(lubridate)
library(lmerTest)
library(psych)Sach Note: “I have exported these documents as examples of the file formatting/data type from the new MetricWire platform so Jackie and I can create a new MW payment script. In total, each export will have 46 downloaded CSVs. The way our new study”ELS T3 Version 2" is set up, there are 43 morning assessments (Daily Sleep x 14, Daily Emotion x 14, Significant Events x 14, and an additional Sleep Diary Survey that counts as one assessment). There are 2 afternoon and evening assessments (Daily Sleep and Daily Emotion). There is one assessment on the last day of the study “ELS Sleep”.
All of the morning surveys take the format of:
1596828076222_Daily Sleep 8_00AM Survey - Day 1.csv
1596828162532_Daily Emotion 8_00AM Survey - Day 1.csv
1596828203346_Significant Events 8_00AM Survey - Day 1.csv
[The above three surveys are identical for the 13 other day assessments, except the day number changes (e.g., 1596828076222_Daily Sleep 8_00AM Survey - Day 2.csv). The number at the front will also probably change w/ each export)
1596828262046_Sleep Diary.csv [this captures sleep diary for all 14 days]"
2 Read in MW sleep data
sleep_files <- Sys.glob(paste("~/Box/Mooddata_Coordinating/ELS_RDoC/At Home Components/3. MetricWire/Survey Monitoring/ELS_T3_Workspace_Version_2/Exported_data/","*","_Daily Sleep 8_00AM Survey - Day *.csv", sep = ""))
sleep_file_exp1 <- "~/Box/Mooddata_Coordinating/ELS_RDoC/At Home Components/3. MetricWire/Survey Monitoring/ELS_T3_Workspace_Version_1/Exported Data/1619543546838_Daily Sleep 6_45am Survey.csv"3 cleaning MW_daily_sleep from the new MW version
Exports survey in as separate days - thus, building for loop to read in those files
sleep_data_exp_v2_df <-
data.frame(
Identifier = NA,
ELS_ID = NA,
Session = NA,
Survey_start_date = NA,
triggerdate = NA,
sleep_trigger_time = NA,
DailySleep_hrs = NA,
DailySleep_satisfaction = NA,
DailySleep_hrs_rec = NA
)
for (s in 1:length(sleep_files))
{
sleepFile = Sys.glob(paste(sleep_files[s],sep=""))
day = unlist(strsplit(sleep_files[s],'/'))[12]
print(day)
sleep_exp2 = read_csv(sleepFile) %>%
rename(
User_ID =
`User Id`,
Session =
`Last Name`,
ELS_ID =
`First Name`,
triggerdate =
`Trigger Date`,
sleep_trigger_time =
`Trigger Time`,
DeviceOS =
`Device OS`,
Survey_start_date =
`Survey Started Date`,
DailySleep_hrs =
`How many hours did you sleep last night?`,
DailySleep_satisfaction =
`How restful or satisfying was your sleep last night?`
) %>%
mutate(
User_ID = as.factor(User_ID)
) %>%
dplyr::select(-c(`Response Id`,
`Submission Location`,
`Trigger Id`,
`Device Id`)
) %>%
mutate(
DailySleep_hrs = as.factor(DailySleep_hrs)
) %>%
mutate(
DailySleep_hrs_rec =
recode_factor(DailySleep_hrs,
"5hr or less" = 5,
"6 hr." = 6,
"7hr." = 7,
"8hr." = 8,
"9hr. or more" = 9)
) %>%
mutate(
DailySleep_hrs_rec =
as.character(DailySleep_hrs_rec)
) %>%
mutate(DailySleep_hrs_rec =
as.numeric(DailySleep_hrs_rec)
)
sleep_exp2_select <-
sleep_exp2 %>%
dplyr::select(ELS_ID,
User_ID,
Session,
Survey_start_date,
triggerdate,
sleep_trigger_time,
DailySleep_hrs,
DailySleep_satisfaction,
DailySleep_hrs_rec)
sleep_data_exp_v2_df <- bind_rows(sleep_data_exp_v2_df, sleep_exp2_select)
}## [1] "1626121823171_Daily Sleep 8_00AM Survey - Day 1.csv"
## [1] "1626121922431_Daily Sleep 8_00AM Survey - Day 2.csv"
## [1] "1626122001771_Daily Sleep 8_00AM Survey - Day 3.csv"
## [1] "1626122103031_Daily Sleep 8_00AM Survey - Day 4.csv"
## [1] "1626144501790_Daily Sleep 8_00AM Survey - Day 5.csv"
## [1] "1626147916204_Daily Sleep 8_00AM Survey - Day 6.csv"
## [1] "1626147996586_Daily Sleep 8_00AM Survey - Day 7.csv"
## [1] "1626148086702_Daily Sleep 8_00AM Survey - Day 8.csv"
## [1] "1626148256141_Daily Sleep 8_00AM Survey - Day 9.csv"
## [1] "1626153690338_Daily Sleep 8_00AM Survey - Day 10.csv"
## [1] "1626153788932_Daily Sleep 8_00AM Survey - Day 11.csv"
## [1] "1626154102413_Daily Sleep 8_00AM Survey - Day 12.csv"
## [1] "1626154259380_Daily Sleep 8_00AM Survey - Day 13.csv"
## [1] "1626154392230_Daily Sleep 8_00AM Survey - Day 14.csv"sleep_data_exp_v2_df <-
sleep_data_exp_v2_df %>%
drop_na(ELS_ID)4 cleaning MW_daily_sleep from the older MW version
Exports survey as one csv for all days
sleep_exp1 <- read_csv(sleep_file_exp1)
sleep_exp1_clean <-
sleep_exp1 %>%
rename(
User_ID =
`User Id`,
Session =
`Last Name`,
ELS_ID =
`First Name`,
Survey_start_date =
`Survey Started Date`,
triggerdate =
`Trigger Date`,
sleep_trigger_time =
`Trigger Time`,
DeviceOS =
`Device OS`,
DailySleep_hrs =
`How many hours did you sleep last night?`,
DailySleep_satisfaction =
`How restful or satisfying was your sleep last night?`
) %>%
mutate(
User_ID = as.factor(User_ID)
) %>%
dplyr::select(-c(`Response Id`,
`Submission Location`,
`Trigger Id`,
`Device Id`)
) %>%
mutate(
DailySleep_hrs = as.factor(DailySleep_hrs)
) %>%
mutate(
DailySleep_hrs_rec =
recode_factor(
DailySleep_hrs,
"5hr or less" = 5,
"6 hr." = 6,
"7hr." = 7,
"8hr." = 8,
"9hr. or more" = 9)
) %>%
mutate(
DailySleep_hrs_rec =
as.character(DailySleep_hrs_rec)
) %>%
mutate(DailySleep_hrs_rec =
as.numeric(DailySleep_hrs_rec)
)
sleep_exp1_select <-
sleep_exp1_clean %>%
dplyr::select(ELS_ID,
User_ID,
Session,
Survey_start_date,
triggerdate,
sleep_trigger_time,
DailySleep_hrs,
DailySleep_satisfaction,
DailySleep_hrs_rec) %>%
drop_na(ELS_ID)4.1 binding dataframes
sleep_data_exp_v1and2_df <- bind_rows(sleep_exp1_select, sleep_data_exp_v2_df)4.1.1 pulling out only “c”
sleep_data_exp_v1and2_df_c <-
sleep_data_exp_v1and2_df %>%
mutate(
porc =
ifelse(str_detect(ELS_ID, "c"),
"C",
"P"
)
)
sleep_data_exp_v1and2_df_c <-
sleep_data_exp_v1and2_df_c %>%
filter(
porc == "C"
)
sleep_data_exp_v1and2_df_c_clean <-
sleep_data_exp_v1and2_df_c %>%
mutate(
ELS_ID =
str_remove_all(ELS_ID,
"[c]")
) %>%
filter(
ELS_ID != "Jalyn",
ELS_ID != "jalyn",
ELS_ID != "sah",
ELS_ID != "vanessa"
) %>%
mutate(ELS_ID =
as.numeric(ELS_ID)
) %>%
dplyr::select(
ELS_ID,
User_ID,
Session,
Survey_start_date,
triggerdate,
sleep_trigger_time,
DailySleep_hrs,
DailySleep_satisfaction,
DailySleep_hrs_rec
) %>%
mutate(
ELS_ID =
as.factor(ELS_ID),
User_ID =
as.factor(User_ID)
)4.1.2 parsing dates
sleep_data_exp_v1and2_df_c_clean_date <-
sleep_data_exp_v1and2_df_c_clean %>%
mutate(
triggerdate = dmy(triggerdate),
surveystartdate = dmy(Survey_start_date)
)4.1.3 creating day and week variables
# creating day and week variables
sleep_data_exp_v1and2_df_c_clean_day <-
sleep_data_exp_v1and2_df_c_clean_date %>%
mutate(
day = day(triggerdate),
wday = wday(triggerdate, label = TRUE),
ELS_ID = factor(ELS_ID)
)
# day num
sleep_data_exp_v1and2_df_c_clean_dayorder <-
sleep_data_exp_v1and2_df_c_clean_day %>%
group_by(ELS_ID) %>%
mutate(
dayorder = order(triggerdate)
)4.1.4 wkday or wkend
sleep_data_exp_v1and2_df_c_clean_dayorder2 <-
sleep_data_exp_v1and2_df_c_clean_dayorder %>%
mutate(
week =
ifelse(
wday == "Sun" |
wday == "Sat",
"wkend",
"wkday"
)
) %>% # keeping ELS_ID 120 and 199 with april enrolled date
filter(
!User_ID == "60e888ca2c155f07c1e85a3b",
!User_ID == "6022082d8d32f32f9848def8"
) %>%
mutate(
User_ID = factor(User_ID)
)5 Reading in response rate (RR) for daily sleep survey
mw_sleep_rr_fp <- "~/Box/Mooddata_Coordinating/1_Lab_Coordinating/Users/JackieSchwartz/Dissertation/mw_daily_sleep_rr.csv"
mw_sleep_rr <- read_csv(mw_sleep_rr_fp)
mw_sleep_rr_clean <-
mw_sleep_rr %>%
dplyr::select(
ELS_ID, Identifier, Enrolled, sleep_resp_rate
) %>%
rename(User_ID = Identifier) %>%
mutate(
ELS_ID = as.numeric(ELS_ID),
ELS_ID = factor(ELS_ID),
dup = duplicated(User_ID)
) %>% # keeping ELS_ID 120 and 199 with april enrolled date
filter(
!User_ID == "60e888ca2c155f07c1e85a3b",
!User_ID == "6022082d8d32f32f9848def8"
) %>%
mutate(
User_ID = factor(User_ID)
)
## how many subjects are we starting out with?
sum_N <-
sleep_data_exp_v1and2_df_c_clean_dayorder2 %>%
group_by(ELS_ID) %>%
summarise(
n = n()
) # 150
# merging with broader df
sleep_data_exp_v1and2_df_c_clean_dayorder_rr <-
left_join(
sleep_data_exp_v1and2_df_c_clean_dayorder2,
mw_sleep_rr_clean,
by = c("ELS_ID", "User_ID")
)5.1 dealing with not enough or missing data
if there are fewer than 3 days of data, I’m removing the subject
# how many days per individual?
emasleep <-
sleep_data_exp_v1and2_df_c_clean_dayorder_rr %>%
group_by(ELS_ID) %>%
mutate(
data_num_obs =
max(dayorder)
) %>%
drop_na(DailySleep_satisfaction) %>%
drop_na(week)
## how many subjects are we starting out with?
sum_N <-
emasleep %>%
group_by(ELS_ID) %>%
summarise(
n = n()
) # starting off with 150
# RR for the full df
describe(emasleep$sleep_resp_rate) # 0.21 to 1## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 1256 0.77 0.23 0.86 0.8 0.21 0.21 1 0.79 -0.82 -0.51 0.01# filter those who have more than 3 days
emasleep_filter <-
emasleep %>%
group_by(ELS_ID) %>%
filter(
data_num_obs > 2
) %>%
filter(
dayorder < 15 # no one had more than 15 days
) %>%
mutate(
Session = str_remove(Session, "ELS-")
)
describe(emasleep_filter$sleep_resp_rate) # 0.2142857 to 1## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 1254 0.77 0.23 0.86 0.8 0.21 0.21 1 0.79 -0.82 -0.5 0.01## how many subjects can we analyze?
sum_N <-
emasleep_filter %>%
group_by(ELS_ID) %>%
summarise(
n = n()
) # 1356 data viz
emasleep_filter %>%
ggplot(aes(dayorder, DailySleep_satisfaction,
group = ELS_ID)) +
geom_line(alpha = 1/2) +
facet_wrap(~ELS_ID) +
theme_minimal() +
theme(
strip.text.x = element_blank()
) +
labs(
x = "Day",
y = "Sleep Satisfaction"
)
ggsave("daily_mw_sleep_traj.png", width = 7, height = 7, dpi = 400)emasleep_filter %>%
ggplot(aes(dayorder, DailySleep_hrs_rec,
group = ELS_ID)) +
geom_line(alpha = 1/2) +
facet_wrap(~ELS_ID) +
theme_minimal() +
theme(
strip.text.x = element_blank()
) +
labs(
x = "Day",
y = "Number of Hours"
)
ggsave("daily_mw_sleep_hrs_traj.png", width = 7, height = 7, dpi = 400)6.0.1 distribution of weekday vs. weekend and association with sleep sat and hrs
level_key <- c(wkday = "weekday", wkend = "weekend")
emasleep_filter <-
emasleep_filter %>%
mutate(
week = factor(week),
week = recode_factor(week,!!!level_key),
dayorder = as.numeric(dayorder),
DailySleep_hrs_rec = as.numeric(DailySleep_hrs_rec)
)
emasleep_filter %>%
ggplot(
aes(x = week, y = DailySleep_satisfaction, fill = week)
) +
geom_violin(alpha=0.5, color= "black") +
geom_boxplot(width=0.1, color = "grey", alpha=0.5) +
scale_fill_manual(values = c("#FFA07A","#6B8E23")) +
theme_classic() +
labs(x = "Weekday or Weekend", y = "Sleep Satisfaction")
ggsave("wkdayvsweeknd_sleepsat.png", width = 7, height = 6)
summary(lmer(scale(DailySleep_satisfaction) ~ scale(dayorder) +factor(week) + (1 + scale(dayorder)|ELS_ID), data = emasleep_filter))## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: scale(DailySleep_satisfaction) ~ scale(dayorder) + factor(week) +
## (1 + scale(dayorder) | ELS_ID)
## Data: emasleep_filter
##
## REML criterion at convergence: 3191.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.1074 -0.5611 0.0777 0.5806 3.0870
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## ELS_ID (Intercept) 0.40772 0.6385
## scale(dayorder) 0.01622 0.1274 0.27
## Residual 0.59088 0.7687
## Number of obs: 1254, groups: ELS_ID, 135
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -0.03055 0.06165 141.70494 -0.496 0.6210
## scale(dayorder) -0.01857 0.02776 72.10588 -0.669 0.5056
## factor(week)weekend 0.12635 0.05177 1126.19864 2.441 0.0148 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) scl(d)
## scal(dyrdr) 0.173
## fctr(wk)wkn -0.197 0.075emasleep_filter %>%
ggplot(
aes(x = week, y = DailySleep_hrs_rec, fill = week)
) +
geom_violin(alpha=0.5, color= "black") +
geom_boxplot(width=0.1, color = "grey", alpha=0.5) +
scale_fill_manual(values = c("#FFA07A","#6B8E23")) +
theme_classic() +
labs(x = "Weekday or Weekend", y = "Subjective Sleep Duration")
ggsave("wkdayvsweeknd_sleephrs.png", width = 7, height = 6)
summary(lmer(scale(DailySleep_hrs_rec) ~ scale(dayorder) + factor(week) + (1 + scale(dayorder)|ELS_ID), data = emasleep_filter))## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: scale(DailySleep_hrs_rec) ~ scale(dayorder) + factor(week) +
## (1 + scale(dayorder) | ELS_ID)
## Data: emasleep_filter
##
## REML criterion at convergence: 3147.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.96224 -0.58190 0.04184 0.62208 2.66292
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## ELS_ID (Intercept) 0.425088 0.65199
## scale(dayorder) 0.008174 0.09041 -0.13
## Residual 0.569567 0.75470
## Number of obs: 1254, groups: ELS_ID, 135
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -0.06952 0.06219 141.20964 -1.118 0.26551
## scale(dayorder) 0.03972 0.02523 72.67905 1.574 0.11981
## factor(week)weekend 0.14983 0.05073 1130.74100 2.953 0.00321 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) scl(d)
## scal(dyrdr) 0.019
## fctr(wk)wkn -0.193 0.075sleep_week <-
emasleep_filter %>%
group_by(ELS_ID, week) %>%
summarize(
n = n()
)## `summarise()` has grouped output by 'ELS_ID'. You can override using the `.groups` argument.emasleep_filter_spread <-
emasleep_filter %>%
spread(week, DailySleep_satisfaction)
emasleep_filter %>%
ggplot(
aes(
x = dayorder,
y = DailySleep_satisfaction,
color = week
)
) +
geom_line(aes(group = ELS_ID, color = week), alpha = .3) +
geom_smooth(method = "loess", se=FALSE, size = 2) +
theme_classic() +
scale_x_continuous(
name = "Day",
limits = c(1, 14),
breaks = seq(1, 14, 1)
) +
scale_y_continuous(
name = "Sleep Satisfaction",
limits = c(1, 100),
breaks = seq(1, 100, 10)
) +
scale_color_manual(values = c("#FFA07A","#6B8E23"))## `geom_smooth()` using formula 'y ~ x'
ggsave("sleepsat_byDay_byWkdayWkend.png", width = 7, height = 6)## `geom_smooth()` using formula 'y ~ x'emasleep_filter %>%
ggplot(
aes(
x = dayorder,
y = DailySleep_hrs_rec,
color = week
)
) +
geom_line(aes(group = ELS_ID, color = week), alpha = .3) +
geom_smooth(method = "loess", se=FALSE, size = 2) +
theme_classic() +
scale_x_continuous(
name = "Day",
limits = c(1, 14),
breaks = seq(1, 14, 1)
) +
scale_y_continuous(
name = "Sleep Satisfaction",
limits = c(5, 9),
breaks = seq(5, 9, 1)
) +
scale_color_manual(values = c("#FFA07A","#6B8E23"))## `geom_smooth()` using formula 'y ~ x'
ggsave("sleephrs_byDay_byWkdayWkend.png", width = 7, height = 6)## `geom_smooth()` using formula 'y ~ x'6.0.2 write
emasleep_filter <-
emasleep_filter %>%
dplyr::select(-Survey_start_date)
write_csv(emasleep_filter, "~/Box/Mooddata_Coordinating/1_Lab_Coordinating/Users/JackieSchwartz/Dissertation/0_MW_Act_Demo_Descriptives/MW_daily_sleep_data_longform.csv")7 sleep instability?
given the huge amounts of variability in sleep satisfaction, the mean may not be an accurate representation computing the rmssd as per Koval et al., 2013; Jahng et al., 2008 The RMSSD is measured as the square root of the average of the squared differences between affect at measurement i and i + 1 (Schoevers et al., 2020)
ssd_df <-
emasleep_filter %>%
dplyr::select(
ELS_ID, surveystartdate, dayorder, DailySleep_satisfaction, DailySleep_hrs_rec
) %>%
group_by(ELS_ID) %>%
summarize(
# step 1: computing successive difference
sleep_succ_diff = DailySleep_satisfaction - lag(DailySleep_satisfaction),
sleep_hrs_succ_diff = DailySleep_hrs_rec - lag(DailySleep_hrs_rec),
# step 2: computing square of each diff
sleep_sq_succ_diff = sleep_succ_diff^2,
sleep_hrs_sq_succ_diff = sleep_hrs_succ_diff^2
) %>%
ungroup()## `summarise()` has grouped output by 'ELS_ID'. You can override using the `.groups` argument.mssd_df <-
ssd_df %>%
group_by(ELS_ID) %>%
summarize(
# step 3: averaging the squared differences
mean_sq_succ_diff = mean(sleep_sq_succ_diff, na.rm = TRUE),
mean_sq_succ_diff_hrs = mean(sleep_hrs_sq_succ_diff, na.rm = TRUE),
# step 4: computing square root of the average
rmssd_sleep = sqrt(mean_sq_succ_diff),
rmssd_sleep_hrs = sqrt(mean_sq_succ_diff_hrs)
) %>%
mutate(ELS_ID = factor(ELS_ID)) %>%
ungroup()7.0.1 computign the mean and dispersion (sd)
mw_sleep_red <-
emasleep_filter %>%
group_by(ELS_ID) %>%
summarize(
dailysleep_sat_mean = mean(DailySleep_satisfaction, na.rm = TRUE),
dailysleep_sat_sd = sd(DailySleep_satisfaction, na.rm = TRUE),
dailysleep_hrs_mean = mean(DailySleep_hrs_rec, na.rm = TRUE),
dailysleep_hrs_sd = sd(DailySleep_hrs_rec, na.rm = TRUE)
) %>%
mutate(ELS_ID = factor(ELS_ID)) %>%
ungroup()7.0.1.1 joining dataframes
mw_sleep_red2 <-
left_join(
mw_sleep_red,
mssd_df,
by = "ELS_ID"
)
mw_sleep_red3 <-
left_join(
mw_sleep_red2,
mw_sleep_rr_clean,
by = "ELS_ID"
)
# joining the df with survey start date
emasleep_filter_select <-
emasleep_filter %>%
dplyr::select(
ELS_ID, surveystartdate
) %>%
group_by(ELS_ID) %>%
mutate(
startdate =
min(surveystartdate)
) %>%
dplyr::select(ELS_ID, startdate) %>%
unique() %>%
mutate(ELS_ID = factor(ELS_ID)) %>%
ungroup()
mw_sleep_red4 <-
left_join(
emasleep_filter_select,
mw_sleep_red3,
by = "ELS_ID"
) %>%
dplyr::select(-User_ID, -Enrolled, -dup)7.1 summary stats
mw_sleep_red_summary <-
mw_sleep_red4 %>%
summarize(
n = n(),
dailysleep_sat_mean_sum = mean(dailysleep_sat_mean),
dailysleep_sat_sd_sum = sd(dailysleep_sat_mean),
dailysleep_hrs_mean_sum = mean(dailysleep_hrs_mean),
dailysleep_hrs_sd_sum = sd(dailysleep_hrs_mean)
)
write_csv(mw_sleep_red_summary, "mw_sleep_red_summary.csv")7.1.1 write
write_csv(mw_sleep_red4, "~/Box/Mooddata_Coordinating/1_Lab_Coordinating/Users/JackieSchwartz/Dissertation/0_MW_Act_Demo_Descriptives/MW_daily_sleep_data_reduced.csv")